Object Recognition software
detect objects in side-scan sonar and video feeds
Object recognition (OR) is an advanced new feature for EvoLogics Sonobot and UUVs.
The AI-based system is an extra module that runs directly onboard the vehicle and analyses raw side-scan sonar or video camera output.
Objects of interest are detected and visually highlighted in the operator’s control software onshore - all live during the mission
A cloud-based ecosystem around the new OR system provides users with regular updates and new detectable object classes. It also allows uploading user datasets for the system to be trained for new object types upon request.
.gif)
EXPANDING RANGE OF DETECTABLE OBJECT CLASSES
Missing persons
Unexploded ordnance
Fish shoals
Fishing nets
Anchors
Antiques
Corals
Shipwrecks
Plane wrecks
Sunken vehicles
Archeology
Custom markers
OBJECT RECOGNITION - KEY FEATURES
Real-time object recognition
-
Raw side-scan sonar or video data is analyzed on-the-fly onboard the vehicle
-
Optimized performance - the system detects objects in less than a second
-
Neural network algorithms that process sonar data can handle large data volumes in real time, and run on dedicated embedded hardware
Expanding range of object classes
-
Use our existing objects library
-
Cloud-based service to train the system for new objects
-
Add your own objects - perform scans and send your data for AI training
Compatible with all EvoLogics vehicles
-
Sonobot 3 uncrewed surface vehicle
-
Sonobot 5 uncrewed surface vehicle
-
Uncrewed underwater vehicles
(coming soon) -
Object recognition as hard- and software upgrade for existing Sonobot USVs
Cloud-based services
-
Regular updates for the on-board software module
-
Create new object classes - submit your scans for analysis and AI training
-
Share your objects or keep for private use
OR for video feed
-
The same hard- and software can process video feeds from onboard cameras
-
Particularly useful for UUVs for extra-detailed surveys and monitoring
“It takes training and experience to recognize object shapes in a sonar picture, and less experienced operators will often miss cues such as contours and sonar shadows. You have to look from multiple angles and normally make several runs at different headings. Thus, a survey can be very time-consuming and tiring for the operator who is often working under pressure, outdoors, in poor weather.
We developed a high-end object recognition system that runs on the vehicle itself and directly analyses the raw data – as cloud-based computing is not available in open water. We had to build neural network algorithms that work on sonar and video data, can handle huge data volumes in real-time, and run on the dedicated embedded hardware. Then, any neural network must be trained to be effective and we have invested a lot of time to develop pre-trained models for the objects, relevant to our key customers. We made sure the system comes with pre-trained object classes and is ready to go out-of-the-box”,
- Philipp Bannasch, Team leader sensor integration at EvoLogics.
Object recognition was first offered to select Sonobot customers in 2020 and over the following months went through extensive testing and optimization.
After the test run, our object recognition with state-of-the-art hard- and software is now available embedded into our commercial surface and underwater robots.